Gemini 3.1 Pro: Pricing and Visual Coding

Gemini 3.1 ProGoogle's frontier model in Cursor for work that combines images and code, especially UI implementation and visual code analysis. Press Enter for the full definition. is Google's Pro model in Cursor. On Cursor's card it is medium for speed and cost and frontier for intelligence. Its card leads with image handling: the model processes images alongside code, which Cursor calls out for UI and UX implementation from mockups, frontend development and visual code understanding. That gives you a more specific reason to choose it than the frontier label does.

Cursor's page still calls 3.1 Pro Google's latest model. Cursor now also lists Gemini 4 ArgonGoogle's newest Gemini model in Cursor at $2 per million input tokens and $10 per million output. Press Enter for the full definition., which its docs call Google's frontier Gemini model, and Google's own model list calls Argon its flagship reasoning model. Of the two Cursor pages, only 3.1 Pro's mentions images, and neither compares the models.

Use 3.1 Pro when the image is part of the evidence. A screenshot showing a responsive failure, a design mockup with measurable spacing, or a visual reference paired with an existing component gives the model something concrete to compare. The model can still do ordinary repository work, but another option may be cheaper when the task has no visual input or large-context need.

Give the model the visual source, the route or component involved, and an explicit acceptance condition. Point to the exact region of a screenshot when you can. Ask it to inspect the implementation before editing, then verify the result in the browser with a new capture. The task becomes a visual comparison backed by code, instead of a prompt asking for something to feel polished.

1. 1Attach the mockup or screenshot and identify the state, viewport, and component it represents.
2. 2Name the route and any files already known to own the UI.
3. 3State measurable constraints such as dimensions, alignment, content order, or interaction behavior.
4. 4Ask the agent to edit, run the application, and inspect the result in the browser.
5. 5Review the final screenshot, diff, responsive states, and automated checks before accepting the change.

Visual understanding doesn't guarantee visual correctness. A screenshot can't show keyboard behavior, loading states, reduced motion or a layout at every width, so add those checks when they matter. Keep a person in the final review for brand judgment. Gemini can make the first pass and test what the browser exposes, but it can't infer a design-system policy nobody wrote down.

Cursor lists 200k as the model's standard context window and 1M as its maximum. The higher ceiling can support a broad repository analysis, several design artifacts or a long tool-driven session. It shouldn't be the default input size, though. Unrelated context can hide the file that matters, and it raises cost before the agent writes a line of code.

Start with the feature slice and expand when a search result or dependency points outward. If the task needs more than 200k, record why. Cursor says input and output pricing increases when input exceeds that threshold, so crossing it changes both the amount of context and the rate applied to the request.

Cursor bills Gemini 3.1 ProGoogle's frontier model in Cursor for work that combines images and code, especially UI implementation and visual code analysis. Press Enter for the full definition. as a third-party model, outside the included pool on The X SubscriptionThe subscription Cursor is sold through: one plan for Grok, Cursor and X, with Free, Plus, Premium, Super and Ultra tiers, per-seat Teams, a weekly included pool for Cursor models shared with Grok, and third-party models billed at API list price from prepaid credit. Press Enter for the full definition.. Its standard rates are $2 per million input tokens and $12 per million output, with cache reads at $0.20. The cache-write cell is blank, so treat that rate as unpublished rather than free.

To use it on The X SubscriptionThe subscription Cursor is sold through: one plan for Grok, Cursor and X, with Free, Plus, Premium, Super and Ultra tiers, per-seat Teams, a weekly included pool for Cursor models shared with Grok, and third-party models billed at API list price from prepaid credit. Press Enter for the full definition. you need Premium or above, and usage comes out of prepaid credit; Plus and Free have no third-party models. On a legacy plan you need Pro or higher, where Gemini 3.1 ProGoogle's frontier model in Cursor for work that combines images and code, especially UI implementation and visual code analysis. Press Enter for the full definition. draws on the Other ModelsOn legacy Cursor plans, the monthly pool for supported third-party models from providers such as Anthropic, Google and OpenAI, separate from Cursor Models. On The X Subscription third-party models are billed from prepaid credit instead. Press Enter for the full definition. pool. Cursor StartAn India-only legacy individual plan at ₹649 a month centered on Cursor's own models, with cloud and mobile access but without Other Models, Auto, Bugbot, Automations or the Agent SDK; Cursor names Plus as the comparable plan on The X Subscription. Press Enter for the full definition., which has only the Cursor ModelsOn legacy Cursor plans, the monthly usage pool for Cursor's own models (Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5), kept separate from the third-party Other Models pool. The X Subscription replaces it with one weekly pool. Press Enter for the full definition. pool, can't use it.

Cached input costs a tenth of the fresh-input rate, so a stable thread that reuses the same project context can cost much less than repeated fresh sessions. Switching models breaks the cache.

On Teams and Enterprise, Cursor's separate Token Rate applies on top, because Gemini 3.1 ProGoogle's frontier model in Cursor for work that combines images and code, especially UI implementation and visual code analysis. Press Enter for the full definition. isn't on the excluded list.

Choose 3.1 Pro when the task needs frontier-rated intelligence or the visual-coding strengths on its model card. Choose Gemini 3.8 FlashGoogle's current speed-tier model in Cursor at $0.75 per million input tokens and $3.75 per million output; it succeeds Gemini 3.7 Flash. Press Enter for the full definition., the Flash model Cursor now recommends, when throughput and output cost matter more and the reasoning demand is bounded. Flash costs $0.75 input and $3.75 output against Pro's $2 and $12, and both list cache reads at a tenth of fresh input.

The older Gemini 3.7 FlashGoogle's previous speed-tier model in Cursor, hidden by default and succeeded by Gemini 3.8 Flash at the same published rates. Press Enter for the full definition. has the same rates as 3.8 Flash but is now hidden by default. The 3.7 Flash page covers it for teams that still pin it.

Let evidence from the result settle it. Run the same bounded task in fresh sessions, then compare the screenshot or artifact, final diff, checks, follow-up corrections, elapsed time and usage. A small price difference can disappear after one failed attempt, and a higher rate earns nothing when the cheaper model's result is as good.

Inside Cursor, Gemini 3.1 ProGoogle's frontier model in Cursor for work that combines images and code, especially UI implementation and visual code analysis. Press Enter for the full definition. can search files and folders, use web search, read and edit files, run shell commands, control the browser, generate images, ask clarifying questions and fetch rules. It reads images when the surface provides them. Tool availability still depends on the current product, approvals, sandbox, credentials and team policy.

Give high-impact tools a clear boundary. A browser test and a screenshot are evidence, but a production deployment or an external write is a separate action that needs its own authorization, whatever tools the model can call. For visual tasks, store the input image and the final capture with the review. The pair makes the model's interpretation checkable later and gives a regression report a stable reference. Without both, a reviewer compares the current screen against a remembered mockup, or against a chat description that may omit the detail that drove the edit.

A team default should match where the work starts. If most tasks begin from screenshots and design files, Gemini's image handling gets used. If nearly every task is plain backend code, that strength sits idle while the team pays a Pro-tier output rate.